In Reply We fully agree that observational studies should be interpreted cautiously and with noncausal explanations thoroughly considered.
James pointed out that the risk estimates for 8 or more cups of coffee per day vs nondrinking were 1.30 (95% CI, 1.17-1.45) in age- and sex-adjusted models, and 0.86 (95% CI, 0.77-0.95) after further multivariable adjustment.1 This change was largely because of adjustment for tobacco smoking, because smokers were more likely to drink coffee than nonsmokers. For this reason, it is important to adjust our models for tobacco smoking, and the availability of detailed information on tobacco use was a strength of the cohort.